I am the boss of me: The executive function of self-awareness in 3- and 4-year-olds

215  Download (0)

Full text



I am the boss of me:

The executive function of self-awareness in 3- and 4-year-olds

Josephine Ross

A thesis submitted for the degree of Doctor of Philosophy, University of Stirling, December 2008




“Some people are scared of children, but they shouldn’t be, they’re just little people with some peculiarities………much like women”

Dr. Robin Campbell. Lunch conversation, 2007.

“I made you out to be a kind of Einstein-like figure, so you don't have much to live up to. Let your hair grow and buy a false moustache.”

Dr. Robin Campbell. Email correspondence 11th July 2008.

Robin, your questionable advice lured me into academia, and your questionable references keep me here. Without question, without you, this thesis would not exist.




This thesis was supported by an ESRC award (PTA-030-2004-00444). Other than the indelible mark of my second supervisor Dr. Robin Campbell, the bulk of acknowledgement is due to my principle supervisor Dr. James Anderson. Jim provided me invaluable inspiration and support for all aspects of the research, whilst recognising my need to think I am my own boss. The same can be said of my husband, John.

Many thanks are due to Stirling Council, and the nurseries who participated in this research. Particular thanks are due to nursery-leaders Aileen Schmidt and Sarah Thorburn and all of the children in Stirling University Playgroup. Every experiment began, and ended, with them.

Thanks also to my friend and colleague Dr. Marina Wimmer, who could always make me a year wiser in the world of academia, and at least a year stupider in the world of schnapps.

Finally, thank you to my friends and family, who didn’t always know what I was doing, but always seemed to think I could do it.




I declare that the work undertaken and reported within this thesis is my own and has not been submitted in consideration of any other degree or award.




Ross, J. Anderson, J.R., & Campbell, R. (2008). It wasn’t me: Self-awareness and self-regulation in preschool children. Paper presented at the British

Psychological Society Developmental Section Conference, Oxford Brookes.

Ross, J., Anderson, J.R., & Campbell, R.N. (2007). Self-representation and self-conservation in the event memories of pre-school children. Paper presented at the




The current research explored the thesis that cognitive self-recognition might have an executive function in 3- and 4-year-olds. Although it is well established that children recognise themselves in mirrors by the end of infancy, the cognitive and behavioural impact of this capacity has yet to be elucidated. Experiments 1 to 6 showed that preschool children could form and maintain a cognitive link between the self and external stimuli, as a result of which, self-referent stimuli were given mnemonic priority. Experiments 4 to 8 indicated that in tasks involving self-recognition, 3- and 4-year-olds’ ability to process other-referent stimuli was compromised by self-focus. Finally, Experiments 9 and 10 demonstrated that mirror self-recognition increased preschoolers’ tendency to self-regulate, leading them to behave in line with socially accepted standards. Together, these experiments provide novel evidence to confirm that cognitive self-recognition has a role in preschoolers’ performance on tasks requiring memory, attention, inhibition, and planning. This implies that when salient, the self may become the ultimate executer of behaviour. By observing 3- and 4-year-olds’ differential processing of self- and other-referent stimuli we infer the existence of a functionally active, self-reflective agent. Moreover, the role of the self is temporally extended, influencing children’s cognition and behaviour in the past (Experiment 1 to 3), present (Experiments 4 to 8) and future (Experiments 9 to 10). This implies that preschool children may have developed the foundations necessary to build the experience of personal identity.



Table of Contents

i. Title page ii. Dedication iii. Acknowledgments iv. Declaration v. Dissemination vi. Abstract

1. The Self in development………....1

1.1 The development of objective self-reflection………2

1.2 The current thesis………..11

1.3 Methodological note………..15

2. Self-reflection and the self in the past I……….17

2.1 Experiment 1………..23

3. Self-reflection and the self in the past II………..…….38

3.1 Experiment 2………..43

3.2 Experiment 3……….……….58

4. Self-reflection and the self in the present I………...70

4.1 Experiment 4……….…….76

4.2 Experiment 5……….…….85

4.3 Experiment 6………..96

5. Self-reflection and the self in the present II………...….103

5.1 Experiment 7………104

5.2 Experiment 8a………..117

5.3 Experiment 8b………...…..125

6. Self-reflection and the self in the future I………...135

6.1 Experiment 9………136

7. Self-reflection and the self in the future II………..154

7.1 Experiment 10………..157

8. The executive self in development………174

8.1 Is “I” the boss of “me” in preschoolers?……….….….174

8.2 Suggestions for future research……….179

8.3 Concluding remarks………181



1. The Self in development

Most simplistically described as the belief “I am me, I was me, I will be me”, personal identity is a universally experienced, yet under researched, phenomenon. In making the claim “I am me, I was me, I will be me” we refer to distinct aspects of the self. A distinction can be made not only between the self in the past, present and future, but between the agentive “I” and the descriptive “me”. Indeed, William James (1890) was the first to make this distinction, breaking self-knowledge into various categories (material, social, spiritual), and referring to the “I” self as the keeper of this knowledge, greater than the whole of its parts. Modern researchers have followed suit, seeking to understand self-awareness by categorising various aspects of the self (for example see Duval & Wicklund, 1972; Neisser, 1988; Lewis 1991). However, the sum of these parts has largely been left to philosophical enquiry. Moreover, in modern research, the Jamesian “I” is often equated with the self perceived subjectively, contrasting this to objective self-knowledge. However, agency is experienced both subjectively and objectively. It requires explicit self-reflection to abstract from the sum of parts the claim for “I”. Likewise, subjective experience arises from many aspects of objective knowledge (consider the knowledge “I am charitable”).

Perhaps understandably, psychologists have preferred to avoid this overlap, studying the self in strict dichotomy. However, over a century since James’ (1890) dissection of the self, attempts to put the self back together again are arguably long overdue. This is regrettable as James (1890), and subsequently Mead (1934),


2 offered a relatively simple empirical route to the study of the elusive “I”. They suggested that agentive experience of personal identity might be measured by focusing on the consequences, as opposed to the content, of self-reflection. The idea here is that in observing the cognitive or behavioural impact of self-reflection, we infer the existence of a reflective agent. The aim of the current thesis is to use this method to explore the ontogeny of self-reflection, and by inference, the experience of personal identity. The present chapter begins by reviewing evidence for the onset of the capacity for objective self-reflection. Although self-reflection is both a subjective and objective process, discussion of a full sense of personal identity cannot begin until the capacity to explicitly acknowledge one’s agency is proven. Having assessed this evidence, the specific aims and methodology of the research to follow is introduced.

1.1 The development of objective self-reflection

Supporting a precocious capacity for self-recognition, newborn infants discriminate their own cries from those of others, failing to show contagion of distress when exposed to their own, rather than another infant’s pre-recorded cry (Martin and Clark, 1982; Dondi, Simion, Caltran, 1999). Moreover, by 3 months of age, infants appear to discriminate their own image from the images of others. Field (1979) demonstrated that 3-month-olds show decreased social and cardiac responses to their own mirror-image versus the equivalent visual stimulus of a peer. Further, Bahrick & Watson (1985), and Legerstee, Anderson & Shaffer (1998) show that 3- and 5-month-old infants are already habituated to their own image, preferring to view a live video image of a novel peer. In support, Legerstee, Anderson & Shaffer


3 (1998) and Rochat & Striano (2002) report that at least by 4 months, infants make significantly more social responses (vocalising and smiling) to live representations of other people, even when contingency is controlled for through mimicry.

However, children do not behave as though they have a cognitive understanding that they are the object reflected in the mirror until the end of infancy. This aspect of awareness is measured by the mirror mark test of self-recognition (independently developed by Gallup, 1970 for use in comparative research, and Amsterdam, 1972 for use with infants). In this test, children are surreptitiously marked (classically with rouge) in a visually inaccessible area (such as the forehead). When a mirror is introduced, children are expected to take appropriate mirror guided action, reaching for, or trying to remove the mark. This behaviour indicates that the child has inferred a relationship between the mirror image and themselves. In other words, they have cognitively identified themselves as an object in the environment. The finding that children typically fail the mark test of mirror self-recognition under the age of 18 to 24 months is robust (see Anderson, 1984 for an early review; later papers include: Asendorf & Baudonniere, 1993; Asendorf, Warkentin & Baudonniere, 1996; Vyt, 2001; Howe, Edison & Courage, 2003; Courage, Edison & Howe, 2004; Lewis & Ramsay, 2004; Nielsen, Dissanayake & Kashima, 2003; Nielsen & Dissanayake, 2004; Bard, Todd, Bernier, Love & Leavens, 2006; Nielsen, Suddendorf, & Slaughter, 2006).

Importantly, younger infants do not fail the mark test due to an inability to follow task requirements. Prompts drawing attention to the marked area, or asking


4 children to wipe the mark with a tissue, do not alter behaviour (Howe, Edison & Courage, 2003; Courage, Edison & Howe, 2004). Likewise, younger infants perform well when asked to attend to directly visible marks on their mother’s face (Lewis & Brooks-Gunn, 1979), a doll’s face (Bard et al, 2006; Asendorf, Warkentin & Baudonniere, 1996), or their own hand (Nielsen, Dissanayake & Kashima, 2003). Comparing infants with and without prior exposure to mirrors, Priel & de Schonen (1986) showed that although locating other objects using the mirror was related to previous experience of reflective surfaces, self-directed behaviour was not. In doing so, Priel & de Schonen (1986) supported the cross-cultural validity of the test. Although there appear to be cultural variations in the onset of mirror self-recognition within the 18 to 24 month window, later work has supported this conclusion (Keller, Yovisi, Borke, Kärtner, Jensen, & Papaligoura, 2004).

Priel & de Schonen’s (1986) results imply that prior experience of mirrors is not necessary for contingency detection between self and mirror image. However, in populations where mirrors are common, it is not clear if contingency detection between self and image precedes mirror self-recognition. Lewis and Brooks-Gunn (1979) reported that 8-month-olds spent significantly longer attending to a live image of themselves than a delayed self-image, or a pre-recorded image of another child. Likewise, Field (1979) reported that, although reacting more positively to an image of a peer, 3-month-old infants spent longer looking at their mirror image. However, as noted, several other researchers report that infants prefer images of others to contingent images of themselves (Bahrick & Watson, 1985; Legerstee, Anderson & Shaffer, 1998; Rochat & Striano, 2002). Although the direction of bias


5 differs, these studies imply that very young infants distinguish displays on the basis of contingency or familiarity. However, in a recent longitudinal sample of 9- to 24-month-old infants, Nielsen, Dissanayake & Kashima (2003) failed to find a significant preference in either direction prior to 18 months. Tracking preferences individually, infants began to prefer their own image to that of a peer (as opposed to looking equally at both images) only in the session where they first demonstrated mirror self-recognition as measured by the mark test.

A less ambiguous measure of contingency detection is provided by object search studies; however, the results remain equivocal. Bertenthal & Fischer (1978) and Bigelow (1981) demonstrated that prior to mirror self-recognition, infants are able to use the mirror to guide their search for objects; for example, reaching up to a hat held above their head, or turning around to fetch a toy. However, more recently, Vyt (2001) and Courage, Edison & Howe (2004) have shown that the ability to use reflective surfaces to infer the location of objects is variable in onset, sometimes preceding and sometimes following mirror self-recognition. Perhaps as a result, using the mirror to guide searches for objects appears statistically unrelated to mark directed behaviour. Nevertheless, in support of a role for contingency detection, identification of self in the mirror typically precedes identification of the self in static photographs by a few months (Lewis & Brooks-Gunn, 1979; Brooks-Gunn & Lewis, 1984; Johnson, 1982; Courage, Edison & Howe, 2004). This implies that physical representation of own features can, at least briefly, be dissociated from cognitive identification arising from proprioceptive feedback.


6 Perhaps the clearest evidence for the importance of contingency is provided by Povinelli, Landau & Perilloux (1996), who demonstrated that 2-year-old children fail non-contingent mark tests of self-recognition. To manipulate contingency, Povinelli, Landau & Perilloux (1996) showed children photographs or videos depicting a marking event which had taken place 3 minutes earlier. Despite labelling these images as self-referent, the majority of 2-year-olds failed to reach for the mark until a contingent stimulus (the mirror) was introduced. Further, it was not until the age of 4 years that the majority of children exhibited mark directed behaviour. Povinelli, Landau & Perilloux (1996) interpret this result as implying that 2- and 3-year-olds cannot objectively connect the experience of past and present selves. In support, 2- and 3-year-olds appear to be aware of the capacity for photographs and videos to reflect other aspects of reality; they can be trained to use these stimuli to guide their search for objects (Suddendorf, 2003; Troseth, 2003; Skouteris, Spataro & Lazaridis, 2006; Skouteris & Robson, 2006; Skouteris, Boscaglia, & Searl, 2007). However, implying that the developmental lag in delayed self-recognition is not as wide as originally proposed, this training facilitates 3-year-olds’ mark directed behaviour (Skouteris, Spataro & Lazaridis, 2006; Skouteris & Robson, 2006).

Even with video-guided training, 2-year-olds fail the delayed test (Skouteris, Spataro & Lazaridis, 2006; Skouteris & Robson, 2006; Skouteris, Boscaglia, & Searl, 2007). Nevertheless, it is not clear that 2-year-olds’ failure is due to a lack of self-awareness. Note, when children use mirrors and delayed videos to search for objects, they have no existing representation of the location of the object prior to the “clue” being provided. On the contrary, Zelazo, Sommerville & Nichols (1999)


7 argue that assuming the existence of an internal self-representation, the mark test requires children to process conflicting representations. Specifically, in both immediate and delayed mark tests, children should experience conflict between an internal self-representation (not marked) and an external self-representation (marked), causing them to take action. Indeed, 2-year-olds’ successful labelling of self-image in the delayed mark test implies that they match featural aspects of the display to an internal self-representation1. However, in the absence of kinaesthetic matching, it appears that the external self-representation is not sufficiently convincing to induce a change in belief, from not marked, to marked. Put simply, this result implies that 2-year-olds value their own self-perception over the novel perception provided by the video or photograph. Although this implies that 2-year-olds may value subjective feedback over objective reasoning when making decisions, it is difficult to characterise their obstinate stance (I am/was not marked) as a failure of self-conservation.

Moreover, there is evidence to suggest that 2-year-olds mistrust video self-representations even when they are live. As implied by the above literature review, it is generally assumed that live video footage is an adequate mirror substitute. However, Povinelli, Landau & Perilloux (1996; Experiment 3) found that only 60% of 3-year-olds passed the mark test in this medium. In response to this, and similar findings (Johnson, 1982; Vyt, 2001), Suddendorf, Simcock & Nielsen (2006) directly compared mirror self-recognition and live video self-recognition in a


For contingent mark tests one need not assume that the youngest children hold a mental representation of their features; however, once self-labelling and self-recognition in photographs is demonstrated, this would appear a valid assumption.


8 sample of 2 to 3-year-olds. Strikingly, although 90% of 2-year-olds passed the mirror test, only 35% passed a mark test using a live video. Tallying with the revised pass mark for the delayed test, success in the live video test did not match that of the mirror mark test until the age of 3 years. This result was replicated by Skouteris, Boscaglia, & Searl, (2007). Although Demir & Skouteris (2008) demonstrated that 2- and 2.5-year-olds’ ability to pass the live mark test could be improved by training, performance was not brought to ceiling. This result implies that video-based tasks put younger children at a disadvantage even when contingency is held constant.

Perhaps as a result of such concerns, an alternative task for self-awareness in which self-reflection is fully internalised has recently been revived. Inspired by the observations of Piaget (1953/1977); Geppert & Kuster (1983) and Bullock & Lutkenhaus (1990) tested whether young children sitting on a mat appreciated that their body weight was an obstacle when attempting to hand the mat to the experimenter. They found that passing this task ontogenetically (Bullock & Lutkenhaus, 1990) and longitudinally (Gepper & Kuster, 1983) preceded mirror self-recognition. More recently, Moore, Mealiea, Garon & Povinelli (2007) have developed a new apparatus, designed to provide a less familiar (and so less easily solved) problem. In this task, children are placed on a rug which is attached to the axle of a shopping trolley, and encouraged to push the trolley towards their mother. However, in order for the trolley to move, children first have to step off of the rug. To 15-month-olds’ evident frustration, the need to remove one’s body weight is not appreciated until 18 to 21 months of age. As might be expected, Moore et al (2007)


9 found that passing this task strongly correlated with mirror self-recognition. However, unlike mirror self-recognition, which at least in later stages involves feature matching, this task does not imply accurate knowledge of the body. Brownell, Zerwas, & Ramani (2007) demonstrated that the majority of 18- to 26-month-olds make body representation errors, such as attempting to put on dolls’ clothes, sit on dolls’ chairs, or squeeze through spaces which are too small.

Despite recent controversy concerning video self-representations, it is commonly accepted that mirror self-recognition is indicative not only of embodied or featural self-awareness, but a wider sense of self-reflection. This conviction is supported by concurrent developments in linguistic and affective domains. For example, from around 18 months of age children begin to refer to themselves using their own name, and as early as 20 months show systematically correct usage of first- (“I, me, my, mine”) and second-person (“you, yours”) pronouns (Brown, 1973; Bates, 1990; Hay, 2006). Implying that linguistic self-reference is premised on objective recognition, children who pass the mirror mark test use more self-other differentiation in language than non self-recognisers (Lewis & Ramsay, 2004; Courage, Edison & Howe, 2004). Moreover, children’s verbal labelling of the mirror image typically lags slightly behind nonverbal behavioural indicators (Amsterdam, 1972; Bertenthal & Fischer, 1978; Harter, 1983; Pipp, Fischer & Jennings, 1987; Bard et al, 2006).

Acting on Amsterdam’s (1972) observation of self-admiring and coy mirror behaviour in 21- to 24-month-olds, Lewis, Sullivan, Stranger & Weiss,


10 (1989) confirmed that embarrassment (averted gaze with smile, blushing, facial touching) typically occurred for self-recognisers, but not non self-recognisers, both in front of the mirror and in public exposure situations (for example, being asked by the experimenter to sing). Children’s empathetic reactions to others’ distress (as measured by sad facial expressions, prosocial helping, sharing, and comforting behaviours) have also been repeatedly linked to the onset of mirror self-recognition (Johnson, 1982; Bischof-Kohler, 1991; Zahn-Waxler, Radke-Yarrow, Wagner & Chapman, 1992). These results are notable, as to feel embarrassment or empathy, one must consider oneself as other, i.e. pass an emotional analogue of the mirror test. However, cognitive and emotional consideration of self and other is not complete at age 2 years. Self-conscious emotions involving evaluation of self to a standard (for example, pride, shame) are not established until at least 3 years of age.

For example, Heckhausen (1984 and later Stipek, Recchia & McClintic, 1992) observed in a number of studies that following success in a competitive task 3- and 4-year-olds (but not 2-year-olds) looked toward their competitor, stretched their body and arms upwards and displayed positive affect. Following failure, the children no longer made eye contact, their body shrunk downwards, and they displayed negative affect. These reactions are consistent with the pro-typical expressions of pride and shame. Further, Lewis, Alessandri & Sullivan (1992) demonstrated that by the age of 3 years, children show more pride when succeeding on difficult than easy tasks, and the converse for failing and shame. This confirms that children’s reactions are not simply based on positive and negative outcomes. More recently Kochanska, Gross, Lin & Nichols (2002) reported that the level of


11 self-recognition reflected in language and behaviour at 18 months correlated with “guilty” behaviour (averted gaze, body tension, distress) following a staged mishap in the laboratory at 22 and 33 months. Mother’s reports of naturally occurring guilty reactions (averted gaze, body tension, distress, seeking reparation) increased during this period.

1.2 The current thesis

The above literature review ends with the culmination of the capacity to reflect on oneself, instrumentally (mark directed behaviour/removing body as obstacle), linguistically (self-referent language), and emotionally (self-conscious and self-evaluative emotions). This end point is justified, not only theoretically, but by necessity. Between the ages of 2- and 4-years, the research reviewed is largely all that is available in experimental psychology2. In introducing the six empirical chapters in the current volume we will encounter some further research that is relevant to our aims. However, in no case does this existing research offer an established view of the early development of objective self-awareness beyond mirror self-recognition. Moreover, beyond mark directed behaviour or exposure emotions, the direct consequences of preschoolers’ self-reflection are yet to be determined. For this reason, the aims of the current thesis are warranted not only by the need to consider the self as a whole entity, but by the need to elucidate the ontogenetic development of the agentive self.


Research concerning the development of children’s self-concepts taken from self- and parental report is not reviewed here due to confound between self-expression and increasing vocabulary skills. However, see Eder (1990) for evidence to suggest that children as young as 3 years can provide consistent reports of their typical temperament and behaviour.


12 A series of ten experiments will be reported, in turn, testing the extent to which preschoolers’ cognition and behaviour imply the subjective and objective belief “I was me”, “I am me”, and “I will be me”. A clear distinction between these aspects of self-awareness, as for subjective and objective experience of the self, is arguably illusory. However, appreciation of the self as a temporally continuous agent is necessary for a full experience of personal identity. Further, as implied by Povinelli, Landau & Perilloux’s (1996) research, the extent to which children younger than 4 years understand the self as an extended entity is yet to be established. Two key routes to measuring the role of personal identity in cognition and behaviour are adopted. Firstly, if an agent is self-reflective we should expect to see a difference in the cognitive processing of self- as opposed to other-referent stimuli. This hypothesis is tested in reference to children’s encoding and retrieval of events (Chapters 2 to 4), and their inhibition of typical processing responses (Chapter 5). If the agent is not only self-reflective but self-evaluative, we might also expect children to adhere to salient standards as a result of self-focus (Chapters 6 and 7). This hypothesis is tested in direct reference to the predictions of Duval & Wicklund’s (1972) theory of objective self-awareness, one of the earliest theories to consider self-consciousness as a functional phenomenon.

In line with other researchers, Duval and Wicklund (1972) dichotomise awareness, suggesting that focus results in explicit and objective self-awareness, whereas external focus renders self-awareness implicit and subjective. However, contrary to the majority of theorists, Duval & Wicklund (1972) recognise


13 the role of subjective processes in contributing to our experience of objective identity. Specifically, Duval & Wicklund (1972) suggest that objective self-focus typically results in positive or negative affect, dependent on one’s perceived consistency with a salient internalised standard for self. As a result of this self-evaluation, those who judge themselves inconsistent with the standard will either adjust their behaviour to conform, or withdraw from the evaluation-inducing situation. In this way, cognitive and affective equilibrium regarding the self is maintained. Any stimulus which reminds one of the self as an object (for example, mirrors, audiences, personal narratives or reports) will induce self-focused attention, and in turn, self-evaluation. Importantly, Duval & Wicklund (1972) recognise that without subjective identification with the self, the motivation to self-regulate is absent3.

Support for the central tenets of Duval & Wicklund’s (1972) theory is plentiful (for reviews see Gibbons, 1990; Fejfar & Hoyle, 2000; Silvia & Duval, 2001a). However, the earliest work offers the simplest illustration of Duval & Wicklund’s (1972) predictions. Support for the assumption that external stimuli can lead to self-focused attention is neatly demonstrated by Davis & Brock, (1975) (and more recently by Stapel & Tesser, 2001). They asked adults to guess the correct


Higgins’ (1987) self-discrepancy theory offers a similar viewpoint. However, Higgins (1987) predicts different affects dependent on whether the situation dictates that the actual self (self-perceived now) is compared to the ideal self (wishes and aspirations) or the ought self (duties and obligations). However, Duval & Wicklund’s (1972) less fractionated model is better suited to the current aims. Similarly, Carver & Scheier (1998) have introduced a cybernetic feedback model of self-awareness, inspired by and extending objective self-awareness theory. Duval has also contributed to extension of his theory, publishing ‘Self-awareness and causal attribution theory’ (Duval & Silvia, 2001) to deal with complex behaviour predictions dependent on the attribution of responsibility. However, as the basic chain of events remains unchanged, and the aim was to detect functional self-reflection rather than model it, Duval & Wicklund’s (1972) original and well supported model was considered the most appropriate frame of reference.


14 pronoun when reading a foreign language; those exposed to a mirror interpreted significantly more pronouns as being personal (I, me, my) than those in the control group. Support for Duval & Wicklund’s (1972) idea of the behavioural consequences of objective self-awareness is clearer still. Carver (1975) found that participants who claimed to oppose punishment as a method of learning gave fewer electric shocks to others in the context of an experiment than those who did not oppose punishment. The converse also held. Crucially, though, this effect was only significant when a mirror was present. In other words, self-standards had a functional impact on behaviour only when attention was self-focused. Similarly, Diener & Wallbom (1976) found that whereas 71% of undergraduates cheated on an anagram task when seated in a room without a mirror, only 7% did so when the mirror was present. These experiments suggest that self-focus promotes adherence to behavioural standards, the breaches of which are known to be the antecedents of self-conscious emotion (guilt, shame, embarrassment).

Despite over 30 years of research refining Duval & Wicklund’s (1972) theory, and considerable interest in moral development (see Eisenberg, 2000 and Kochanska & Aksan, 2004 for review), little attention has been given to the ontogeny of these effects. Thus, a primary aim for the current research was to determine if preschool children behave as though self-aware as predicted by Duval & Wicklund’s (1972) model. In addition to testing the morally valanced aspects of Duval & Wicklund’s (1972) theory (Chapters 6 and 7), attempts are made to determine if the introduction of an opportunity for explicit self-focus impacts upon non emotive aspects of cognitive processing (Chapters 3 and 4). This was judged


15 important as the developmental lag between the onset of exposure emotions and evaluative emotions suggests that the capacity to focus and the capacity to self-evaluate are dissociable. Following full discussion of the motivations, methods, and results of each experiment, the current volume culminates with a summary of what this research implies for preschoolers’ subjective and objective experience of self-reflection. Given the lack of prior research in this area, suggestions for future research are a recurrent theme, revisited and extended in the final chapter. Although characterising the sum of self-awareness available to preschoolers was never likely to be achievable in the context of a single project, the key contribution of the present thesis is to offer novel methodology which has the potential to elucidate this complex phenomenon.

1.3 Methodological note

In total, 771 preschool children were recruited, with parental consent, from thirteen mixed demographic nurseries in Stirling, Scotland. In most cases testing took place in a separate room within the nursery building. Each study used a roughly equal number of boys and girls. Although children as young as 2.5 years participated in the research, the target age-range for comparison was between 3 and 4 years. This demographic was targeted as by the age of 3 years, typically developing children’s ability to self-recognise linguistically (necessary in Chapters 2 and 3), in mirrors (necessary in Chapters 4 to 7) and photographs (necessary in Chapters 4 and 5) can be safely assumed.


16 The research used a variety of methods, described in full for each experiment. Means (M) and standard deviations (SD) are reported for relevant data (in Figures standard deviations are shown as error bars). All statistical analyses are two-tailed and where possible exact p values are reported. For quantitative data, multivariate and mixed models analyses of variance (ANOVAs) are used to investigate main effects and interactions; in both, age-group is used as a between-subjects factor, together with experimental conditions where appropriate. Post-hoc Bonferroni tests are used to explore between-subjects effects, and split sample analyses for mixed level interactions. For all ANOVAs partial eta squared (ηp2) is presented as an estimate of effect size. This value can be interpreted following Cohen’s (1969) criteria: ηp2 > 0.2 large, > 0.1 medium, and > 0.05 small. Occasionally, Pearson’s correlations are used to explore effects where age-group analyses are not appropriate, or to compare performance across tasks. Where relevant, above chance performance is assessed using probability calculations and one-sample t-tests. Finally, for qualitative data presented in Chapters 6 and 7, non-parametric signed-rank tests (Friedman’s k related samples, Wilcoxon’s paired samples) are used for within-subject comparison, and Pearson’s chi-square analyses are used to explore the distribution of behaviour within conditions.



2. Self-reflection and the self in the past I

An important component of personal identity is recognition of the continuity between past and present selves. One of the ways in which adults express this knowledge is through autobiographical event narratives. In describing their role in past experiences, adults are expressing their ongoing identification with a past self. For this reason the reporting of autobiographical memories can be considered an explicit expression of self-conservation. However, autobiographical memories cannot be reduced to external expression. To access and maintain information pertaining to one’s involvement in a past event requires an internal, autobiographically organised system. It is arguably this system, as opposed to its product (a narrative), which is at the root of personal identity. Following this reasoning, Chapters 2 and 3 aim not to elicit early autobiographical event narratives, but to experimentally assess the extent to which the self has an active role in event memory in the preschool years.

Complementary to cross-cultural studies emphasising the pervasiveness of autobiographical life narratives (see Fivush & Haden, 2003), there is a large body of research suggesting a general mnemonic advantage for material which has been encoded as relevant to the self. The subject-performed task effect (SPT) refers to an established memory bias for action statements that have been acted out relative to statements recited verbally, or witnessed being performed by others (see Engelkamp 1998, for review). This effect is thought to occur due to the greater depth of processing involved in an event in which one actively engages. The cognitive


18 equivalent, a memory bias for adjectives judged to be self-descriptive relative to those processed without reference to the self, is known as the self-reference effect (SRE) (see Symons & Johnson, 1997 for meta-analysis).

The SPT effect and SRE are well established in adults; however, relatively few studies have addressed the ontogeny of these phenomena. This neglect is regrettable as the emergence of a mnemonic bias toward self-related material is relevant to the development of both self-awareness and autobiographical organisation in memory. The SRE is thought to be based on the organisational properties of a highly elaborated self-concept (Symons & Johnson, 1997). By contrast, the SPT effect is often considered to be based on lower-level proprioceptive feedback (for example, Engelkamp, 1998). According to this interpretation, physically experienced events are processed at a deeper level than passively experienced events without recourse to cognitive identity. However, there is evidence to suggest that at least when comparing memory for self-performed versus other-performed actions, the SPT effect relies on higher level self-awareness.

Firstly, children with clinical impairments in objective self-awareness fail to show self-reference effects in social paradigms. Millward, Powell, Messer & Jordan (2000) tested free- and verbally prompted recall for target events experienced by autistic and normally developing children on a 25-minute walk. Half of the target events were experienced by the child, and the remainder by a companion. After they had returned from the walk, Millward et al (2000) found that normally developing 5- to 6-year-olds remembered and expressed more information relating to events


19 they had actively experienced. However, autistic children of the same verbal mental age showed the opposite bias, recalling more information about the experiences of their companion. Confirming that asymmetrical performance is not traceable to a lack of proprioceptive engagement, Summers & Craik (1994) reported that both normally developing and autistic children (each with an average mental age of 5 years) showed the same magnitude of bias for recognition of self-performed versus self-verbalised tasks. Consequently, it has been argued that although children are capable of benefiting from agency on a proprioceptive basis (for further evidence see Williams & Happe, 2008), they fail to take (or at least to capitalise on) the cognitive perspective of the self in memory (Powell & Jordan, 1993; Russell & Jarrold, 1999).

Secondly, in typically developing children, the bias for self- versus other-performed actions is tempered by the level of cognitive identification with “other”. Baker-Ward, Hess & Flannigan (1990) asked 5- to 8-year-old children to take turns performing 21 actions with objects and later to recall the actions made. Children were questioned immediately after the play session, and again three weeks later. Some of the children were also questioned about the event in the intervening weeks. Baker-Ward, Hess & Flannigan (1990) found that regardless of timeframe, repeated questioning, or age, children showed a bias for correct recall of actions they had performed relative to those they had observed being performed by randomly selected classmates. However, in a second study, Baker-Ward, Hess & Flannigan (1990) showed that the mnemonic advantage for self-performed actions disappeared when contrasted with memory for actions performed by the children’s most regular


20 playmates. It is possible that self-performed actions were well remembered due to subjective feedback, whereas actions performed by highly familiar others benefited from their association with an elaborated person-concept. However, it is difficult to see how cognitive familiarity could play a role for other’s actions, but not for own.

Millward et al (2000) and Baker-Ward, Hess & Flannigan’s(1999) research confirms that higher-level cognition plays a mediating role when comparing memory for self- and other’s actions. For this reason, investigation of the functional effect of self-reflection on memory need not preclude action-based paradigms. Indeed, the use of physical play seems better suited to developmental research than the complex processing of word lists required by the standard SRE. Nevertheless, researchers using the standard SRE paradigm have had some success in demonstrating an effect for young children. For example, Bennett & Sani (2008) have recently demonstrated a SRE for trait descriptions in a sample of 5-, 7- and 10-year-olds. Children recalled more simple adjectives (clever, friendly, funny, greedy, happy, messy, naughty, noisy, small, and rough) which had been processed with the question “Do you think you are _____?” than those processed semantically (Do you think ___ means the same as ____ ?). Pullyblank, Bisanz, Scott & Champion (1985) earlier reported positive results in a similar study comprised of 7- and 10-year-olds. Like the SPT effect studies reviewed above, neither of these SRE studies uncovered a significant interaction between mnemonic self-bias and age, implying that the effects of self-involvement are functional in children as young as 5 years.


21 Interestingly, Bennett & Sani (2008) found that answering the question “Do you think people in your family are ______?” also led to better recall than semantic processing. This effect, reminiscent of the equalising effects of familiarity found by Baker-Ward, Hess & Flannigan (1990), has been repeatedly found in SRE studies using adult participants. This has led to the criticism that mnemonic self-bias is premised on familiarity, as opposed to self-awareness per se (see Symons & Johnson, 1997). However, social research strongly suggests that identification with a social group (comprised of familiar others) is a major aspect of an elaborated self-concept (see Johnson, Gadon, Carlson, Southwick, Faith & Chalfin, 2002, Bennett & Sani, 2008). In others words, both self-processing and familiar other processing seem likely to involve recourse to cognitive identity. Moreover, it is difficult to see how self-awareness and familiarity could, or should, be separated. The self is by definition, uniquely familiar. Likewise, given the continuous link the self provides between encoding and retrieval conditions, it is arguably in a unique position to facilitate memory. Most importantly, whatever the “specialisation” of the underlying mechanism, the minimum requirement for a cognitive SRE is self-recognition.

Using a new paradigm, especially designed to circumvent the complexity of trait adjectives and semantic judgements, Sui & Zhu (2005) also provide evidence that SREs occur in children as young as 5 years old. Sui & Zhu (2005) presented 4-, 5- and 10-year-old children with pictures showing various objects being pointed to by a generic figure. To manipulate self-referencing this figure was altered to include a photograph of either the child’s own or an unfamiliar child’s face. For each


22 presentation the child was asked to report who was pointing to the object. After a short distraction task, Sui & Zhu (2005) introduced a surprise recall test for the encoding phase. They found that although all age-groups were above chance in monitoring who had pointed to the objects they recalled, only 5-year-olds named significantly more of the objects associated with their self-image than those associated with the other image. Ten-year-olds showed a non-significant bias towards self-referent material, and 4-year-olds a non significant bias in the opposite direction. In a second experiment, Sui & Zhu (2005) demonstrated that when the task demands were sufficiently challenging for 10-year-olds, they also showed a significant SRE.

Sui & Zhu’s (2005) study is perhaps the first to directly assess the impact of self-recognition on memory in children under the age of 5 years. Both Howe, Courage & Edison (2003) and Harley & Reese (1999) found that mirror self-recognition had a positive effect on very young children’s event recall; however, their tests of self-recognition were parallel to, rather than integrated with the memory task. There was no attempt to elicit or measure autobiographical organisation of the to-be-remembered event. In interpretation of their results, Sui & Zhu (2005) suggest that prior to the age of 5 years, children’s self-concepts, although present, may not be sufficiently elaborated to have a functional effect on memory. However, adults report autobiographical details of events occurring between the ages of 3 and 4 years (Pillemer & White, 1989). Importantly, for an adult to remember details of an event as it happened to them as a child, a link between the adult and child self must be maintained. It is difficult to see how this


23 link could persist if the event details were not originally encoded as self-referent. This suggests that the impact of the self on the organisation of event memories begins in the preschool years, and that Sui & Zhu’s (2005) conclusion is open to challenge4.

2.1 Experiment 1

SRE/SPT paradigms appear well suited to investigating the development of autobiographical organisation in memory because they make a functional link between self-reference and memory. Moreover, in contrast to pass/fail mark tests of mirror self-recognition, the magnitude of the SRE/SPT effect is likely to be a relatively sensitive tool to measure developmental change in self-reflection. Due to the preliminary nature of the investigation this potential has yet to be fully explored. To investigate the onset of functional self-awareness in memory, Experiment 1 used a new paradigm which aimed to combine the investigation of SRE and SPT effects using a delayed measure of one week. A delay of one week was chosen to determine if the SRE was maintained in long-term memory (necessary to be considered autobiographical). To manipulate subjective self-reference, children were asked to take turns with the experimenter to perform actions. To manipulate objective self-reference, each action was introduced by one of two cartoon peers; a boy and a girl. Concepts of own gender and age-group are among the first aspects of self-knowledge expressed (Stipek, Gralinski & Kopp, 1990). To further stimulate identification with gender-matched peers, this character was given the same name as



24 the child. This explicit label provides one of the simplest entry points to activate self-recognition. Although our own name is highly (if not uniquely) familiar, it is also inextricably linked, post objective self-awareness, with the idea of “me”.

One motivation for combining SRE and SPT effects in the same study was to ensure that the stimuli were memorable. When physically involved in an event, 5-year-olds have been shown to maintain self-bias in event memory up to three weeks later (Baker-Ward, Hess & Flannigan, 1990). If preschoolers could also maintain a SRE or SPT effect over a delay, this would imply not only a functional role for self-awareness at encoding, but a capacity for “autobiographical” retrieval. In turn, this would imply maintenance of a cognitive link between past and present selves. Another motivation to combine the study of SRE/SPT effects was to contribute new data to discussion regarding the basis of the SPT effect in social paradigms. Finding SRE and SPT effects of different magnitudes or onsets in the same age-range would imply that SRE and SPT effects are functionally distinct. Finally, as noted, in addition to providing children with physical scaffolding for the processing the stimuli, cognitive scaffolding was provided in the form of cartoon figures. An advantage of this method was that the cartoon figures provided a visual record of the actions performed in the encoding session. This allowed children’s free recall of the stimuli to be supplemented with a recognition measure. Such measures are important as young children are likely to be relatively poor at narrative encoding and retrieval of perpetually based events (see Simcock & Hayne, 2002).


25 Method


Forty-five children from three age-groups took part; 15 3-year-olds (M = 36.5 months, SD = 2.9 months, range = 31 - 41 months), 15 3.5-year-olds (M = 44.7 months, SD = 1.5 months, range = 42 – 47 months) and 15 4-year-olds (M = 53.5 months, SD = 4 months, range = 48 - 59 months). Ten additional children (six 3-year-olds, four 3.5-year-olds) were excluded due to failure to follow the procedure for taking turns in action re-enactment.


The experimental stimuli at encoding consisted of an introductory drawing of two preschool children (one male, one female) standing side by side and facing the viewer, and two sets of 18 A6 action cards (including two practise cards) depicting these children performing an action. In each set, half of the actions were depicted by the male, and half by the female. The actions depicted by each character were counterbalanced across sets5. At retrieval two sets of 16 A4 recognition cards were used, the set selected matched the action cards used at encoding. Each recognition card depicted four actions (of the same type) performed by the same actor, one of which had been introduced at encoding. All actions were comparably


An equal number of actions depicted could be categorised as self-directed (acts on self with or without object), other-directed (actions directed away from self, with or without object) and actions with or without an imaginary object. Post-hoc, these action categories appeared to have no influence on performance and so are not discussed in detail.


26 simple to perform and familiar to preschool children. Examples of the stimuli used are shown in Figure 2.1.

Introductory card

i) ii) iii)

Modelled Action cards

i) ii) iii)

Action recognition cards




In the first session children were introduced to the two cartoon characters on the introductory card (order counterbalanced). The default names of the characters were Mary and Bob. However, children were routinely told that the character of their own gender shared their own name. So, for example, for a child named Louisa, the first session would commence: “Today we’re going to see Louisa and Bob do lots of different things then we’re going to take turns acting just like them.” To ensure they understood the procedure, children were shown two practise action cards and instructed: “Let’s have a practise. Here is Louisa clapping, and here is Bob standing on one leg. I am going to clap my hands, just like Louisa (perform action). Can you stand on one leg just like Bob?” (or vice versa). Having successfully completed this practise phase children were praised and reminded only to perform actions when prompted by the experimenter. As noted, children were excluded if they failed to follow this instruction.

In the encoding phase, children were presented with action cards one at a time in random order: “Look! In this picture name is description of action”, and instructed either to perform the modelled action themselves, or watch the experimenter perform it. The procedure was counterbalanced so that child and experimenter each performed half of the actions modelled by the male child, and half by the female child. When all of the action cards had been presented the child was praised for their participation and given a sticker as a reward.


28 One week later, children were re-introduced to the drawings of the actors and reminded of the previous session: “Do you remember last time I was here and we met a little girl called Louisa and a little boy called Bob? Can you remember any of the things we saw them do last time?”. After responding to this free recall question, children were told that they were going to be shown some drawings to help them remember more about the previous session. A total of 16 recognition cards were introduced one at a time in random order. As noted, each recognition card showed one of the actors performing four actions, one of which had been presented at the previous session. After each of the actions on the recognition card was described briefly, the child was asked “Can you remember which of these things we saw X doing last time?” When all of the recognition cards had been presented the child was again praised and given a sticker as a reward.



Only seven children (four 3.5-year-olds, three 4-year-olds) spontaneously reported any of the 16 target actions from the encoding session. Of the children reporting explicit memories of the previous session, four recalled two of the 16 target actions, and three recalled only one. The low incidence of action recall precludes statistical analysis; however, of the 11 actions recalled, the majority (seven) were self-related on both encoding dimensions (performance and model).




On average children recognised just under half of target actions from the encoding session (M = 7.3, SD = 3.4), no children performed at ceiling. Nevertheless, a one-sample t-test indicated that this success rate was significantly greater than the 25% (1 in 4) predicted by chance (t (44) = 6.4, p < 0.01). As shown in Table 2.1, this held true for every group. However, although the older age-groups performed above chance in the recognition of both self and other-related stimuli, 3-year-olds’ performance was above chance for self-related stimuli only.

Table 2.1: Experiment 1 recognition performance split by age-group and encoding dimension

Age-group Recognition

3-year-old 3.5-year-old 4-year-old

Overall M=5.6 (35%), SD=2.6 (t (14) = 2.39, p = 0.03) M=6.6 (41%), SD=3.4 (t (14) = 2.93, p = 0.01) M=9.8 (61%), SD=2.9 (t (14) = 7.7, p < 0.01) Self-performed M=2.9 (36%), SD=1.7 (t (14) = 2, p = 0.05) M=3.5 (44%), SD=1.9 (t (14) = 2.9, p = 0.01) M=5.6 (70%), SD=1.6 (t (14) = 8.9, p < 0.01) Other-performed M=2.6 (32%), SD=1.4 (t (14) = 1.8, p = 0.09) M=3.1 (39%), SD=1.8 (t (14) = 2.3, p = 0.04) M=4.1 (51%), SD=1.9 (t (14) = 4.1,p = 0.01) Picture model self-referent M=3 (37%), SD=1.7

(t (14) = 2.3, p = 0.04)

M=3.3 (41%), SD=1.8 (t (14) = 2.8, p = 0.01)

M=5.4 (68%), SD=1.8 (t (14) = 7.4, p < 0.01) Picture model other-referent M=2.6 (32%), SD=1.4

(t (14) = 1.6 p = 0.1)

M=3.3 (41%), SD=1.9 (t (14) = 2.6, p = 0.02)

M=4.3 (54%), SD=1.5 (t (14) = 5.8, p < 0.01)


30 Age had a significant main effect on the total number of actions recognised (F (2, 42) = 8.03, p = 0.01, ηp2 = 0.28). Specifically, 4-year-old children recognised significantly more target actions than the younger age-groups (post-hoc Bonferroni tests p < 0.01). As shown in Table 2.2, analysis of the effect of age-group on recognition of actions from each category indicated that the 4-year-olds’ advantage was particularly strong for the recognition of self-related stimuli (note ηp2).

Table 2.2: Experiment 1 effect of age-group on recognition, split by encoding dimension

Recognition Between-subjects ANOVA

Main effect of age-group

Post-hoc Bonferroni test

Self-performed F (2, 42) = 9.7, p < 0.01, ηp2

= 0.3 4 year old – 3-year-old p < 0.01 4-year-old – 3.5-year-old p = 0.006 Other-performed F (2, 42) = 2.7, p = 0.07, ηp2 = 0.1 4 year old – 3-year-old p = 0.08

4-year-old – 3.5-year-old p = 0.3 Picture model self-referent F (2, 42) = 8.6, p = 0.01, ηp2

= 0.3 4 year old – 3-year-old p = 0.01 4-year-old – 3.5-year-old p = 0.006 Picture model other-referent F (2, 42) = 4.3, p= 0.02, ηp2

= 0.17 4 year old – 3-year-old p = 0.017 4-year-old – 3.5-year-old p = 0.2

Figure 2.2 shows that proportionately more self-related than other-related actions were recognised on both encoding dimensions. A 2 (self-performed versus other-performed) x 2 (self-referent actor versus other-referent actor) repeated-measures ANOVA, including age-group as a between-subjects variable, indicated a significant mnemonic advantage for both self-performed actions (F (1, 42) = 7.2, p = 0.011, ηp2 = 0.15) and actions modelled by a self-referent actor (F (1, 42) = 4.96, p


31 = 0.031, ηp2 = 0.11). These effect sizes are equivalent to the typical effect size found in Symons & Johnson’s (1997) meta-analysis of SRE effects.

Figure 2.2: Mean number of self and other-related target actions recognised in Experiment 1

Neither self-related bias was found to significantly interact with age-group (SPT effect, F (2, 42) = 1.9, p = 0.15, ηp2 = 0.08; SRE, F (2, 42) = 1.7, p = 0.2, ηp2 = 0.07). Nevertheless, as evident in Table 2.1, 4-year-olds showed a larger self-related bias than the younger age-groups on both encoding dimensions. Moreover, there was developmental progression in the number of individuals expressing a bias: 60% 3-year-olds, 67% 3.5-year-olds, and 100% 4-year-olds showed a self-bias on at least one dimension.

0 1 2 3 4 5 6 7 8 Encoding dimensions M ea n n u m b er o f a ct io n s re co g n is ed

Performer (SPT effect) Picture Model (SRE) _____________________________________________________chance Self-related Other-related * * * significant p < 0.05


32 To investigate the developmental factor further, and to determine the effect size for each age-group, a repeated-measures ANOVA investigating self-bias was run separately for each age-group. These tests indicated that the self-related mnemonic bias reached significance only for 4-year-olds (SPT effect, F (1, 14) = 6.7, p = 0.015, ηp2 = 0.3; SRE, F (1, 14) = 7.7, p = 0.021, ηp2 = 0.3). As shown in Table 2.1, this age-group showed a large self-reference effect on both dimensions. Although in the expected direction, the bias did not reach significance for 3-year-olds (SPT effect, F (1, 14) = 0.3, p = 0.5, ηp2 = 0.02; SRE, F (1, 14) = 0.7, p = 0.4, ηp2 =0.05) or 3.5-year-olds (SPT effect, F (1, 14) = 1.1, p = 0.3, ηp2 = 0.07; SRE, F (1, 14) = 0.04, p = 0.8, ηp2 = 0.003).


Experiment 1 provides evidence of a mnemonic advantage for self-related material in preschool children. By the age of 4 years, children showed a significant bias for recognising stimuli that were related physically (through performance) or cognitively (through nominal/gender matching) to themselves at encoding. This confirms that both SRE and SPT effects have an earlier onset than implied by previous research (Pullyblank et al, 1985; Baker-Ward, Hess & Flannigan, 1990; Summers & Craik, 1994; Millward et al, 2000; Sui & Zhu, 2005; Bennett & Sani, 2008). Only a few children offered free recall of the actions performed at encoding. Nevertheless, of these memories, more referred to past events in which the child had participated nominally or physically than those passively experienced. If the SPT effect was based entirely on subjective feedback, then one might expect it to


33 precede the “objective” SRE in ontogeny. However, Experiment 1 found SPT and SRE effects of an equivalent magnitude. Moreover, the onset of significant effects was delayed until the second half of the 4th year. As the physical component was controlled in the SRE, this result lends some support to Millward et al (2000) and Baker-Ward, Hess & Flannigan’s (1990) suggestion that objective self-recognition contributes to the SPT effect, at least in social situations.

Observation of a developmental lag in the magnitude (though not the direction) of the SRE and SPT effects measured confirms that this paradigm has the potential to measure developmental change. However, it is possible that the developmental change observed in Experiment 1 is not attributable to cognitive elaboration of the self-concept. For example, younger children’s expression of the SRE/SPT effect may have been compromised by task-specific demands at encoding. Qualitative improvements occur in the understanding of external representations between the 3rd and 4th year (Zelazo, Sommerville & Nichols, 1999; DeLoache, 1991). For this reason, older children might have been better able to identify with the to-be-remembered cartoon representation than their younger counterparts. This would result in a relatively strong SRE.

Older children were less likely to need coaching in order to inhibit action performance when it was the other player’s turn. As out-of-sequence action performances would undermine the distinction between self and other-related stimuli, children were excluded from the study if they could not follow this rule. However, it was evident that for many of the younger children, this involved


34 effortful control (for example, sitting on hands); failure to do this resulted in ten 3- to 3.5-year-olds being excluded from the experiment Similar difficulties were reported when 3-year-olds were required to follow the action instructions of one puppet but not another (Jones, Rothbart & Posner, 2003). Importantly, there is evidence to suggest that the younger children’s tendency to “over-participate” in the game may have blurred the distinction between self and other-performed actions, even when suppressed.

It has been established (Foley, Passalacqua & Ratner, 1993; Foley & Rather, 1998; Ratner, Foley & Gimpert, 2002; Foley, Ratner & House, 2002; Sommerville & Hammond, 2007) that when engaged in a “turn-taking” activity, 3- to 5-year-old children internalise the actions of their playmate to the extent that they later claim the latter’s actions as their own (dubbed the “I did it” bias). Importantly, under these collaborative circumstances, Foley & Ratner (1998) report that young children’s memory for stimuli that were previously acted upon personally or by their partner is equivalent. Of current interest, children sometimes show a form of self-reference effect in this paradigm. Sommerville & Hammond (2007) demonstrated that 3- and 4-year-olds’ learning from an activity, as measured by re-enactment of previous actions, increases as their “I did it” bias increases (Sommerville & Hammond, 2007). In other words, preschoolers’ cognitive engagement with an activity appears to improve as a function of their perception of self-performance, even where this label is unwarranted. For this reason, if Experiment 1’s 3-year-olds were prone to judge the actions performed as collaborative, the SPT effect might extend to both self-performed and other-performed actions.


35 We might also find an explanation for the observed developmental differences by considering the task demands at retrieval. Despite efforts to support their memories, the youngest children still appeared to struggle to recognise the task stimuli, performing significantly worse than 4-year-olds. Developmental improvement in recognition memory capacity was expected. However, just as it is imperative that children have an equal opportunity to encode events as self-referent, their opportunity to retrieve material should also be age-appropriate. This was the motivation for the picture recognition test. However, given younger children’s difficulties in interpreting representations (Zelazo et al, 1999; DeLoache, 1991), a more appropriate way to support perceptual memory might have been to introduce re-enactment of the action sequences at recall and recognition. Preschool children are capable of re-enacting more actions from a previous activity than they can express verbally (Smith, Ratner & Hobart, 1987; Ratner, Foley & McCaskill, 2001). However, asking the children to enact actions at retrieval would introduce a cueing bias for self-performed actions. Only these actions would be experienced in the same mode (action) at encoding and retrieval; i.e. only the target item would constitute re-enactment. Even at recall, the instruction “show me an action from last time” is likely to preferentially cue actions that were previously “shown” by the child.

For this reason, although re-enactment is likely to prove useful in assessing the importance of the subjective aspect of the SPT effect (see Mulligan & Horstein 2003), it is unsuited to establishing its natural occurrence. By contrast, children


36 objectively experience the cartoon picture cards in the same mode (visual) at both encoding and retrieval. The pictures led to different performance experiences at encoding, and they were likely to induce a processing bias if the child interpreted one but not the other as being self-referent. However, any systematic bias is traceable to the child (the SRE/SPT effect), not to the encoding specificity of the retrieval cue provided. For these reasons, it seems that picture recognition tests may be advantageous for determining the natural onset of SRE/SPT effects. However, one way to reduce the visual complexity of these tests would be to reintroduce objects involved in actions at encoding, as opposed to action-object models. As objects are more easily labelled than action statements, this might also reduce the linguistic demands of recall. This method gave Summers & Craik (1994) a positive result for 5-year-old children, and is used for 3- and 4-year-olds in the experiments presented in Chapter 3 of this volume.

In addition to the practical challenges of retrieval, there is theoretical reason to expect a developmental lag in long-term SRE between the 3rd and 4th year. As noted in Chapter 1, Povinelli, Landau & Perilloux’s (1996) results imply that 3-year-olds may have difficulty making reference to the self in the past, at least without training (Skouteris, Spataro & Lazaridis, 2006; Skouteris & Robson, 2006). On this account, SRE/SPT effects may not have been present for younger children one week after encoding, not due to a generally unelaborated self-concept, but due to a self-specific retrieval failure. Indeed, not only were 4-year-olds better than younger children at retrieval, this was largely attributable to better recognition of self-related stimuli. However, the observation that children from all age-groups tend